Adequate nutrition throughout early life is fundamental for proper growth, development, and survival but poor nutrition continues to be a major public health problem in developing countries [1]. Globally, an estimated 32% or 186 million and 10% or 55 million children below 5 years of age in developing countries are stunted and wasted, respectively [2]. According to the latest demographic and health survey, 18.8%, 4.7%, and 5.9% of children under 5 years are stunted, wasted, and underweight, respectively, in Ghana [3]. The prevalence of undernutrition in the study region is slightly lower with 17.2%, 4.5%, and 11.0% of the children being stunted, wasted, and underweight, respectively. Undernutrition in children results in life-long consequences, including increased risk of disease, poor cognitive development, lower school performance and increased risk of death in childhood; reduced productivity; and increased risk of chronic disease in adulthood [4,5,6].

The main causes of child undernutrition are poor quality diets which are low in calories and essential nutrients and repeated infections [7]. These are underpinned by poor child feeding practices, unsanitary environment, restricted access to health services, and household food insecurity. “Food security defines a situation in which all people at all times have physical and economic access to sufficient, safe and nutritious food which meets their dietary needs and food preferences for an active and healthy life” [8]. Household food security exists when all household members, at all times, have access to enough food for an active and healthy life. Household food insecurity depicts food unavailability at the household level and can negatively affect food consumption, resulting in reduced dietary variety and nutrient intake and, consequently, malnutrition in children. Consequently, some studies have shown positive associations between household food insecurity and poor child nutritional status [9,10,11] while others have not [12,13,14].

An indicator of food security at the household level is the household dietary diversity (HDD) score which is a simple count of the number of food groups consumed by a household over a period normally the previous 24 h before the survey [15, 16]. In recent years, dietary diversity scores have become increasingly popular because they correlate well with nutrient adequacy and anthropometric indices of children [17], they are perceived to be important to nutrition and health [18], and quantifying food intake in population studies is challenging.

While a number of studies have examined the association between household food security and child nutritional status [9, 12, 19], very few used household dietary diversity as an indicator of household food security. Similarly, there are a number of studies on individual dietary diversity and child nutritional status [20,21,22,23,24], but again very few on household dietary diversity and child nutritional status. We report the findings of a study conducted to evaluate HDD score as a determinant of nutritional status of children (6–36 months) in Wenchi Municipality in Brong Ahafo Region of Ghana.


Study design and area

A population-based analytical cross-sectional study was conducted in Wenchi Municipality of the Brong Ahafo Region in Ghana. The Wenchi Municipality, located in the Western part of the region, is one of the 22 municipalities in the Brong Ahafo Region. It lies within latitudes 7° 30′ South and 8° 05′ North and longitudes 2° 15′ West and 1° 55′ East, covers 1145 km2 of land area, and shares boundaries with Techiman to the West, Kintampo Municipality to the North, Sunyani to the South and Tain to the East. The total population of the municipality is 89,739, out of which 19,657 are children within the ages of 0–59 months and 23,588 are women in fertile age [25]. The municipality is divided into three (3) operational sub-municipalities namely Wenchi, Subinso, and Droboso, and these have populations of 46,194, 22,605, and 29,485, respectively.

Study population and subjects

The study population comprised of pre-school children aged 6 to 36 months and their mothers/caregivers selected from 30 clusters or communities in the Wenchi Municipality. For a mother-child pair to be eligible for the study, they must be willing to participate and the child should not be sick at the time of the survey or should not have had an episode of diarrhea or malaria in the past week before the survey.

Sample size and sampling technique

The sample size was determined using the single population formula n = t2 × p(1 − p)/m2 [26], where n = required sample size, t = confidence level at 95% (standard value of 1.96), p = prevalence of stunting in children under five in Brong Ahafo Region in 2008 (25.0%), and m = margin of error at 5% (standard value of 0.05). This gave 288 which was multiplied by 2 to account for cluster sampling and increased to 590 to take care of incomplete or spoilt questionnaires.

A two-stage cluster sampling procedure was used for the data collection: 30 clusters were selected at the first stage and 20 mother-child pairs in each selected cluster at the second stage. Using probability proportional to size 14, 9 and 7 clusters were drawn from the 3 sub-municipalities Wenchi, Droboso, and Subinso, respectively, using simple random sampling without replacement [27]. From each selected cluster, the mother-child pairs were recruited from households selected using simple random sampling without replacement [28]. Sampling of the households was based on the WHO/EPI cluster survey methodology. We located the center of each selected cluster with the help of a community health volunteer. At the center, a bottle was spun on the ground and all the houses in line with the direction of the bottle top were numbered. One household was selected at random and visited. The second household selected was the one nearest the first household moving in the chosen direction. The procedure was followed until all the 20 households were selected in the selected cluster. For each selected household, only one eligible mother-child pair was selected using simple random sampling if there was more than one.

Data collection methods and tools

The data for the study were collected in September–October 2014 using a semi-structured questionnaire, a standard questionnaire for measuring HDD score [15, 16], and anthropometry. Data were collected on socio-demographic and economic characteristics of parents of sampled children, HDD, child feeding practices, and child anthropometry. The semi-structured questionnaire was designed following literature review [29, 30].

Measurement of household dietary diversity score

The number of different food groups consumed by households within the previous 24 h preceding the survey was investigated through qualitative 24-h dietary recall. As mothers were mostly responsible for the preparation of meals for the households, they were asked to recall all the foods that the household members had consumed in the house in the past 24 h. Mothers were also made to mention all the ingredients used in the preparation of the meals of the households. Based upon these data, it was recorded whether the household consumed food(s) from one or more of the following food groups: (1) cereals, (2) white roots and tubers, (3) vegetables, (4) fruits, (5) meat, (6) eggs, (7) fish and seafood, (8) legumes, nuts, and seeds, (9) milk and milk products, (10) oils and fats, (11) sweets, and (12) spices, condiments, and beverages [15, 16]. Each food group was counted only once irrespective of the number of food items consumed from it. The HDD score could therefore range from 0 to 12 depending upon the number of food groups the family ate from. There is no recommended cut-off value for HDD score, so using the mean HDD score for the study subjects (6) as recommended [16], the HDD score was categorised into two: “Low” HDD score (1 to 5 food groups) and “High” HDD score (6 to 12 food groups).


Anthropometry was used to assess the nutritional status of children. Weight and height or length were measured for each child using standard techniques [31]. A UNICEF electronic scale (SECA 890) for both mother and child was used to weigh children. Children who were 24 months or above were weighed directly standing on the scale in minimal clothing and those below 24 months were weighed being carried by their mothers.

Standing height was measured for children 24 months and above and recumbent length for children less than 24 months, using an infantometer. Both height and length were measured to the nearest 0.1 cm. During the course of the fieldwork, all equipment were checked for measurement accuracy on a daily basis. Age of children was deduced from date of birth taken from the Child Health Record booklet and date on which the measurements were made.

Determination of household wealth index

A household wealth index based on household assets was generated and used as a proxy indicator for socio-economic status (SES) of households. This is based on an earlier concept whereby the sum of dummy variables created from information collected on housing quality, and ownership of household durable goods was used to construct an index [29]. The 14 variables used were radio set, television set, satellite dish, sewing machine, mattress, refrigerator, DVD player, computer, fan, mobile phone, bicycle, motorcycle, animal drawn cart, and car. A household was awarded a score of “1” for each item possessed otherwise a score of “0,” and using principal component analysis, a wealth score was derived for each household and classified into tertile. Each item was counted only once irrespective of the number the household possessed.

Quality assurance

The questionnaires were obtained or first prepared in English, translated into Akan, and then back translated into English to ensure consistency. The questionnaires were also pre-tested in a non-study community to identify and correct inconsistencies, to check clarity of items, ambiguity of language, and feasibility of administration in a community setting. A 3-day training was given to data collectors during which they practiced how to ask the questions and take anthropometric measurements. The data collectors were community health nurses who were familiar with taking weight and height measures of children. The training also covered informed consent procedures, sampling technique, and interviewing skills. During the course of the fieldwork, the questionnaires were checked for completeness on site, and on a daily basis, and the problems identified were rectified.

Statistical analysis

Data analysis was performed with Stata (version 9) and WHO Anthro. Three anthropometric z-scores (height-for-age, weight-for-height, and weight-for-age) were constructed from weight, height, age, and sex data collected on the children comparing with the WHO Child Growth Standards. Children whose height-for-age, weight-for-height, and weight-for-age z-scores were less than − 2 standard deviations were classified as stunted, wasted, and underweight, respectively [32].

Frequencies and percentages were calculated for categorical variables and means and standard deviations for continuous variables. Chi-square and logistic regression analyses were used in assessing the association between exposure and outcome variables. Chi-square was used to explore the crude associations between parents’ and children’s socio-demographic characteristics and HDD score (exposure variable). Again, chi-square test was used to explore the crude association between parents’ and children’s socio-demographic characteristics and stunting (outcome variable) or wasting (outcome variable). In establishing the association between HDD score and child stunting or wasting, factors that were statistically significant in the bivariate tests and were controlled for. The level of statistical significance for all analyses was set at p < 0.05.

Ethical consideration

The study protocol was approved by the Joint Ethics Board of School of Medicine and Health Sciences and School of Allied Health Sciences, University for Development Studies, Tamale. Written informed consent was obtained from mothers before the administration of the questionnaire and taking of anthropometry data. The mothers of the children were assured of anonymity and confidentiality of the study data. They were also informed that participation was voluntary and that they could withdraw at any time during the process if they felt uncomfortable.


Socio-demographic and economic characteristics of mothers

The mean age of the respondents was 27.3 [95% confidence interval (CI) = 26.8–27.8] years, and 58.3% of them were within the age group of 20–29 years (Table 1). About a quarter of the respondents (27.8%) had no formal education, and 28.8% had education up to junior high school level only. The majority of the respondents were married (82.4%) and Christians (78.5%), while most (43.1%) belonged to the Akan ethnic group. In relation to socio-economic characteristics, most mothers of the children (38.0%) and majority of fathers (51.9%) were engaged in farming. There is some interaction between maternal age and marital status, as more than 70% of mothers aged 15–19 years were not married. Further information on socio-demographic characteristics is provided in Table 1.

Table 1 Socio-demographic characteristics of parents and children

Nutritional status of children

The mean age, weight, and height of the children were 18.6 (95% CI = 17.9–19.3) months, 10.4 (95% CI = 10.0–10.9) kg, and 76.2 (95% CI = 75.6–76.9) cm, respectively. Most children (42.4%) were within the ages of 12 to 23 months, and about half were males (50.5%). Table 2 shows information on the nutritional status of children aged 6 to 36 months in the Wenchi Municipality. The mean height-for-age, weight-for-height, and weight-for-age z-scores were − 1.43 (95% CI = − 1.66 to − 1.20), 0.44 (95% CI = 0.27–0.62), and − 0.46 (95% CI = − 0.58 to − 0.34), respectively. The overall prevalence rates of stunting, wasting, and underweight were 38.6% (95% CI = 35.0–43.0), 11.0% (95% CI = 8.5–14.0), and 13.1% (95% CI = 10.3–16.0), respectively. It can be observed that stunting prevalence increased significantly with age (p < 0.001) while the rates of wasting and underweight were relatively stable (p > 0.05). Four hundred and thirty (430) of the children were below the age of 2 years, so we collected information on their breastfeeding and complementary feeding. About 80% (78.8%) of the children under 2 were reported to have been exclusively breastfed, and 74.9% met the criteria for minimum meal frequency. The mean age of starting complementary feeding was 3.1 months.

Table 2 Nutritional status of children aged 6–36 months

Household dietary diversity score

The mean HDD score for the study population was 6.1 (95% CI = 5.9–6.2). Overall, 36.4% of the households were in low HDD score category meaning they consumed foods from at most five (5) food groups in the previous 24 h preceding the survey. Most households consumed 6 food groups (23.7%) but none consumed all the 12 food groups (Table 3).

Table 3 Number of food groups consumed

Association between household dietary diversity and child nutritional status

In order to identify possible confounders to control for in the assessment of the association between HDD score and child nutritional status, we compared the nutritional status of children (stunting and wasting) belonging to low and high DD score categories (Table 4). For stunting, the following characteristics were associated: marital status (p = 0.009), sex of the household head (p = 0.001), child age (p < 0.001), and sex of child (p = 0.019). For wasting, the following characteristics were associated: mother’s age (p = 0.014), mother’s religion (p = 0.010), household wealth index (p = 0.010), and sex of child (p = 0.003). Stunting was not related to maternal age, educational level, religion, ethnicity, or occupation, and wasting was not related to maternal educational level, marital status, ethnicity, occupation, or sex of the household head. Also, wasting was not related to the ages of the children.

Table 4 Bivariate analysis of the association between maternal and child socio-demographic variables and child nutritional status

Again we compared low and high DD households with respect to socio-demographic characteristics of mothers and children (see Additional file 1: Table S1). An association was found between mother’s age (p = 0.001), maternal education level (p = 0.044), sex of the household head (p < 0.001), household wealth index (p < 0.001), or child’s age (p = 0.007) and HDD.

In the adjusted logistic regression analyses, children belonging to low and high DD score households had similar risks of stunting (adjusted OR = 1.18, 95% CI = 0.79–1.76, p = 0.409) and wasting (adjusted OR = 0.97, 95% CI = 0.54–1.73, p = 0.910) which were not significantly different (Table 5).

Table 5 Multivariate analysis of the association between household diversity score and child nutritional status


We sought to explore the association between HDD and child growth indicators in a municipality in Brong Ahafo Region of Central Ghana. A high prevalence of chronic malnutrition is observed, and HDD is poor but it does not appear to influence the growth of young children in this study population.

A high prevalence of stunting (38.6%) was observed among children 6–36 months in this study population. According to the recent Demographic and Health Survey, one in five children under 5 is stunted nationally and 17.2% of children in the Brong Ahafo region have stunting [3]. The high prevalence of stunting estimated is not unexpected given that the study area was selected based on reports of poor child nutrition [33] in order to evaluate potential risk factors and inform health programming. In agreement with the Ghana Demographic and Health Survey [3], we found that stunting increased with age highlighting problems with complementary feeding as the children transitioned from breastfeeding to the family diet. Sub-optimal complementary feeding practices have been reported to associate with poor growth in children [34]. It is observed that more male children than female children are stunted. In an analysis of demographic and health survey data for 10 sub-Saharan African countries, it was found that stunting is more prevalent in boys than girls [35]. The reason for this observation is unclear but may relate to biological mechanisms conferring greater vulnerability on males [35, 36] or female sex preferences of mothers. Whatever the underlying mechanism is, this finding suggests that caregivers need to pay more attention to the nutrition and health of male children. We however found lower levels of stunting for children of currently married mothers compared to those of currently unmarried mothers and for those in male-headed households compared to those in female-headed households. This may relate to the roles played by husbands and fathers such as provision of resources for food, health care, and other necessities in male-associated households that are enjoyed by children in such households [37]. Stunting was however not related to maternal age, educational level, religion, ethnicity, or occupation.

The prevalence of wasting and underweight (11.0% and 13.1%, respectively) are about twice the Brong Ahafo regional figures of 4.5% and 5.9%, respectively [3]. However, children belonging to younger or older mothers, mothers who practice non-Christian religion, and mothers in low/medium wealth index households, and male children are at increased risk of wasting. Childbearing at younger age is a risk factor for adverse mother and child outcomes including undernutrition [38]. This is because younger maternal age is associated with being unmarried, under-educated, and socio-economically disadvantaged, all of which are risk factors of adverse pregnancy outcomes [39]. On the other hand, older maternal age has been associated with increased risk of wasting [40]. The association of wasting to religion and household wealth index may be related to high socio-economic status of parents and associated increased access to food and health care. Wasting was not related to maternal educational level, marital status, ethnicity, occupation, sex of household head, or child age.

Our study did not find an association between HDD and stunting or wasting. Although children belonging to low DD households are 18% more likely to be stunted compared to children in high DD households, this did not reach statistical significance. Household food insecurity primarily results in inadequate dietary intake in children meaning they may not meet their daily energy and nutrient requirements. Secondarily, household food insecurity may be associated with reduced food variety and as a result low dietary diversity. It is therefore to be expected that households with poor food access would have increased risk of child malnutrition but we did not observe this. In a recent study, children living in food secure households were significantly more likely to receive adequate diet compared to children in food insecure households [41].

Several studies have reported associations between individual DD and child nutritional status [22, 23] while others have not [10, 24]. Some studies reported an association between HDD score and stunting or underweight [9,10,11] while others like ours did not [12,13,14]. The lack of association between HDD and child nutritional status in our study area may reflect a lack of association between household diet and child nutrition in the area or our study’s lack of sufficient power to detect this association if it exists. It is observed that food insecure households protect their children from the effects of food insecurity by prioritising their nutrition over that of adults [42]. So mothers may skip meals to enable their children to eat and by so doing protect them from the ill effects of household food insecurity.

Our study has some limitations. Firstly, we used a cross-sectional study design which does not study causal link. Secondly, the use of a 24-h dietary recall for estimating HDD increases the risk of recall bias and may not reflect the usual dietary pattern. Thirdly, we used cluster sampling to recruit the study participants but did not control for clustering in the analysis. Consequently, we computed confidence intervals for estimates instead of standard deviations. Finally, our sample size may have been inadequate to detect an association between HDD and child nutritional status (if it exists) in our study population. Despite these limitations, our study provides some important insights into the relationship between household diet and child nutritional status in Central Ghana.


In conclusion, both stunting in children and low HDD are prevalent in the study area. We did not observe an association between HDD and stunting or wasting in this study population. Interventions aimed at improving child growth and HDD are needed to promote and protect child health.